We design and build apps, websites and AI tools for founders and product teams.

Ideas, developed.

Mobile and web apps, AI automation, machine learning and computer vision. You bring the blurry version; we bring it into focus.

Fixed quote first · A working build every Friday · You own the code · Replies within one working day

Rec 00:00:00:00
12:00 · MiddayDrag, or use the arrow keys: a quarter hour per press, an hour with Shift.
Scroll to develop

Five crafts, one reel.

Five things we build, from first prototype to production — each one lit at its own hour on the studio wall.

Frame 01 / 05TC 00:00:00
0107:00 · Morning

Mobile apps

iOS and Android apps that feel native, work offline and stay quick on a three-year-old phone. Built in Swift, Kotlin or React Native — whichever your product actually needs.

  • Swift
  • Kotlin
  • React Native
  • Expo
Start a mobile app project
0212:00 · Midday

Websites

Marketing sites and web apps that load before the coffee cools. Content your team can edit without calling us, and motion that earns its place.

  • Next.js
  • TypeScript
  • Headless CMS
  • WebGL
Start a website project
yourbrand.com1440px
LCP 0.8s
0315:00 · Afternoon

Machine learning

From messy spreadsheets to models in production: forecasting, recommendations, language. Evaluated honestly, versioned properly and watched after launch.

  • PyTorch
  • scikit-learn
  • MLflow
  • Feature stores
Start a machine-learning project
epoch 000/120loss 2.270val_acc 41.0%
0.00.61.21.82.40306090120ckpt saved ✓
train validation
0417:00 · Afternoon

AI automation

LLM agents and workflows wired into the tools you already use — inbox, CRM, Slack — with a human in the loop wherever a mistake would cost money.

  • LLM agents
  • RAG
  • n8n
  • Python
Start an automation project
New emailtriggerClassifyllm · 0.4sUpdate dealcrmPost to #salesslackDraft replyhuman approves
demo runs 1,284sample data
0522:00 · Night

Computer vision

Detection, tracking, OCR and visual inspection — trained on your footage, not a demo set, and fast enough for edge devices on a factory floor.

  • YOLO
  • OpenCV
  • ONNX
  • Jetson
Start a computer-vision project
CAM_03 · 24 FPSpeople 000   cars 000
detector · int8 · 6.1 mssignal DRIVE

Which one is yours?

Replies within one working day, from someone who'll work on it.

A day withDayframe.

Every project runs on the same clock: a scoped plan, something real every Friday, and a calm launch. Here's the day, compressed.

  1. 09:00

    Discover

    We sit with your team, your data and your users until the problem is boringly clear. You leave with a scope that has real numbers on it.

    Fixed-fee scope · risks · quote

  2. 12:00

    Design

    Prototypes you can tap, not decks you have to imagine — tested with real users before anyone writes production code.

    Clickable prototype

  3. 15:00

    Build

    Short cycles, a working build every Friday, and a shared channel where every decision is visible as it's made.

    Weekly builds

  4. 18:00

    Ship

    Store releases, deploys, data pipelines and the dull checklists that keep launch day quiet.

    Production release

  5. 00:00

    Run & learn

    Models retrain, automations run, dashboards keep watch. We stay on and improve whatever the numbers point at.

    Monitoring · retraining

Every project starts with a conversation.

Ask for a 20-minute call

Three waysto start.

Pick the one that fits where you are now. Prices here are samples, to give a sense of scale; a scope sprint ends with a fixed quote. Not sure which fits? Price your app in a minute.

  • Senior hands only.

    The people on your first call are the people who design and build it. Nothing is passed down to juniors.

  • Something real every Friday.

    A working build you can open on your phone, every week, so problems show up in days, not months.

  • You own all of it: code, models, data.

    It lives in your accounts from day one and is yours once each invoice is paid.

Beforeyou write.

Something else? Ask it in your brief

How much does an app or website cost?

A first version typically costs $15,000–60,000 (a sample range, not a quote), depending on platforms, features and how custom the design is. A two-week scope sprint gives you a fixed quote before you commit, and its fee is credited against the build.

How soon can you start?

Usually within two to four weeks of a signed scope. If something is urgent, say so in your brief and we'll tell you honestly whether we can make it.

Who owns the code, models and data?

You do. Source code, trained model weights, datasets and design files are yours once each invoice is paid, and they live in your accounts from day one.

Will you sign an NDA?

Yes. Send yours or ask for ours; we're happy to sign before you share any details.

What if it doesn't work?

You see a working build every Friday, so problems show up in days, not months. Machine-learning work starts with a small pilot that measures whether the model is good enough before anything is built around it.

What happens after launch?

Every build includes 30 days of fixes. After that you can keep a small team on a monthly plan, or take everything in-house with documentation and a handover call.